#asahi-dev · 2025-07-14 17:44
Lightswordanyone know what "nvme-apple 27bcc0000.nvme: RTKit: syslog message: blog.c:102: New BlogBuf low 27" means? seeing that a bit in dmesg#asahi-dev · 2025-07-13 22:21
jannauit's even used in this way in the kernel. so it might okay in principle#asahi-dev · 2025-07-13 22:14
chaos_princessyes. parent devices should be first in order#asahi-dev · 2025-07-13 22:12
jannauso platform_device_register_full() still works and passes data to the childs due to probe order?#asahi-dev · 2025-07-13 10:40
jannaumight be time to convert to auxilary bus/dev. see nova-core / nova-drm as example#asahi-dev · 2025-07-13 10:32
jannaufor example "of:NdcpT(null)Capple,t8112-dcpCapple,dcp"#asahi-dev · 2025-07-13 10:32
jannauit looks that way because that's how it's build for of devices. it would have the compatible strings appended with an 'C' as prefix if the node had that#asahi-dev · 2025-07-13 10:24
chaos_princessill look into it#asahi-dev · 2025-07-13 10:23
chaos_princessyea, but that does not look like the modalias we'd want#asahi-dev · 2025-07-13 10:23
jannauit changes because of_dev_modalias starts to work with fwnode set#asahi-dev · 2025-07-13 10:18
chaos_princessthat does not sound like it is looking for the right thing#asahi-dev · 2025-07-13 10:15
jannauchaos_princess: the module load issue is wrong modalias. with fwnode set it's "of:NaudioT(null)" instead of "platform:snd_soc_apple_aop" and the module only hast the latter#asahi-dev · 2025-07-13 10:13
jannauI've pushed rebased/squashed branches (asahi-wip and everything rust related)#asahi-dev · 2025-07-13 08:53
chaos_princessno, it is not, i think the "old-style chassis" ones have a slightly different set of endpoints#asahi-dev · 2025-07-13 08:51
jannaudo you if aop las is present on all macbooks? doesn't seem to be on j493#asahi-dev · 2025-07-13 08:38
chaos_princessals should probably be `depend BROKEN` tbh#asahi-dev · 2025-07-13 08:36
jannauyes, I still use it in als but I had hoped I could get rid of the custom error message#asahi-dev · 2025-07-13 08:34
chaos_princessand sure, i will rebuild as module and test it soon-ish#asahi-dev · 2025-07-13 08:34
chaos_princesswell, you need to unwrap the PropertyGuard or what its called so might as well do .required_by()#asahi-dev · 2025-07-13 08:30
jannau.required_by() is quite pointless then#asahi-dev · 2025-07-13 08:29
jannaudo we know the length of the properties? I guess it's in the worst case aop FW version dependent and so we can't avoid the fwnode.property_count_elem()#asahi-dev · 2025-07-13 08:28
jannaupushed the of -> fwnode conversion for aop_als to asahi-wip#asahi-dev · 2025-07-13 08:22
jannauchaos_princess: that seems to break aop subdevice loading completely when built as modules#asahi-dev · 2025-07-13 07:24
jannauah, subtle#asahi-dev · 2025-07-13 07:17
chaos_princessactually filling out fwnode instead of leaving it null#asahi-dev · 2025-07-13 07:17
jannauchaos_princess: what is supposed to fix probe order? I don't see anything in the PR#asahi-dev · 2025-07-12 21:25
chaos_princessok, figured out the aop probe order issue (ofc the issue is that i am a dumbass and did not understand how to register platform devices correctly)#asahi-dev · 2025-07-12 20:14
jannaualloc_device() fails with "XHCI timeout on event type 33..." / "Cannot allocate device context to get SLOT_ID"#asahi-dev · 2025-07-12 20:13
jannauany good idea what could differ between m2 and m2 pro mac mini that breaks the asmedia xhci? pcie and firmware appear to work but u-boot does not detect devices on the m2 mini#asahi-dev · 2025-07-12 14:44
chaos_princessrebased new gpu init, seems to work fine#asahi-dev · 2025-07-12 14:21
chaos_princessso, i guess just steal whatever pci and similar buses do?#asahi-dev · 2025-07-12 14:20
jannauif aop_audio would be probed by driver core the device link between between admac and aop_audio should ensure that aop_audio is probed after admac (if it were working)#asahi-dev · 2025-07-12 14:17
chaos_princessi am not sure i understand, the dma dependency is "sideways" not parent-child, so moving it down wouldn't change it?#asahi-dev · 2025-07-12 14:14
jannau"compatible" for aop_audio, getting rid of `dma-names = "invalid-order-only"` for aop#asahi-dev · 2025-07-12 14:13
chaos_princessdid you mean aop_audio?#asahi-dev · 2025-07-12 14:10
jannaucan we probe aop_admac through normal device probing, i.e. adding a compatible for the audio child node? I think it's guaranteed that the child probes after the parent (could be the other way round). that would at least get rid of the fake "dmas" property in aop#asahi-dev · 2025-07-12 13:18
chaos_princessi know why the aop/admac probe order is kind of a mess, just not sure how to make it less so#asahi-dev · 2025-07-12 13:18
chaos_princesssure, i will look at it soon-ish.#asahi-dev · 2025-07-12 12:34
jannaualso aux device bindings but I'm not sure platform device are expected to use that#asahi-dev · 2025-07-12 12:33
jannaurust has device properties. feel free to replace OF properties with that for aop/sep#asahi-dev · 2025-07-12 12:29
jannauchaos_princess: didn't you look into that at some point?#asahi-dev · 2025-07-12 12:29
jannauI've seen the aop_audio a couple of times fail due of_dma_request_slave_channel() failing since admac wasn't probed yet#asahi-dev · 2025-07-12 12:27
jannauand either the hack to ensure probe order between aop and admac is not working anymore or never did#asahi-dev · 2025-07-12 12:26
jannauknown issues: lockdep complaints about a potentially cyclic lock on asahi probe#asahi-dev · 2025-07-12 12:26
jannauI've pushed a v6.15.6 based kernel to asahi-wip. untagged so far, please test but looks good from my side#asahi-dev · 2025-07-12 10:28
chadmedthe whole 15.x series is totally FUBAR for our purposes #asahi-dev · 2025-07-12 09:22
chaos_princess15.1 had another bug preventing progress#asahi-dev · 2025-07-12 09:20
yukaIt was successfully booting the raw custom kernel objects#asahi-dev · 2025-07-12 09:19
yuka15.1 I think#asahi-dev · 2025-07-12 09:19
yukaYou can always downgrade to the initial mac os version for m4